Classification and Localization of Disease with Bounding Boxes from Chest X-Ray Images

The use of deep learning has become increasingly popular in current biomedical science circles with the recent surge in availability of many types of medical data. One of the most popular machine learning fields within healthcare is computer vision, an area that has achieved success across a variety of datasets and usages. Chest x-rays are the most common type of radiology exam in the world1, but diagnosing one of the possible chest afflictions to the many organs and systems in the chest is a difficult task. The NIH Chest X-ray Dataset2, released in 2017, is one of the largest publicly available X-ray datasets, and has spawned a number of models to predict disease from the x-rays. Most of these models, including the famous CheXNet3, are implemented to predict the binary classification of the presence of each disease.